Analysis and Simulation of Nonlinearities in Noise Attenuation Model for a Diesel Engine Block

Transcription

1 Purdue University Purdue e-pubs Open Access Theses Theses and Dissertations January 2015 Analysis and Simulation of Nonlinearities in Noise Attenuation Model for a Diesel Engine Block Jiajun Cao Purdue University Follow this and additional works at: Recommended Citation Cao, Jiajun, "Analysis and Simulation of Nonlinearities in Noise Attenuation Model for a Diesel Engine Block" (2015). Open Access Theses This document has been made available through Purdue e-pubs, a service of the Purdue University Libraries. Please contact epubs@purdue.edu for additional information.

2 Graduate School Form 30 Updated 1/15/2015 PURDUE UNIVERSITY GRADUATE SCHOOL Thesis/Dissertation Acceptance This is to certify that the thesis/dissertation prepared By Jiajun Cao Entitled ANALYSIS AND SIMULATION OF NONLINEARITIES IN NOISE ATTENUATION MODEL FOR A DIESEL ENGINE BLOCK For the degree of Master of Science in Mechanical Engineering Is approved by the final examining committee: Peter H. Meckl Chair J. Stuart Bolton Gregory M. Shaver To the best of my knowledge and as understood by the student in the Thesis/Dissertation Agreement, Publication Delay, and Certification Disclaimer (Graduate School Form 32), this thesis/dissertation adheres to the provisions of Purdue University s Policy of Integrity in Research and the use of copyright material. Approved by Major Professor(s): Peter H. Meckl Approved by: Anil K. Bajaj / James D. Jones 7/8/2015 Head of the Departmental Graduate Program Date

3 ANALYSIS AND SIMULATION OF NONLINEARITIES IN NOISE ATTENUATION MODEL FOR A DIESEL ENGINE BLOCK A Thesis Submitted to the Faculty of Purdue University by Jiajun Cao In Partial Fulfillment of the Requirements for the Degree of Master of Science in Mechanical Engineering August 2015 Purdue University West Lafayette, Indiana

4 ii ACKNOWLEDGMENTS I would like to express my deepest gratitude to Dr. Peter H. Meckl, my major professor and research supervisor, for providing me with the opportunity to work on this project, and for his invaluable guidance and support during my entire stay at Purdue University. His insight and enthusiasm were priceless in my understanding of the topic and completion of this thesis. I would also like to thank Dr. J. Stuart Bolton and Dr. Gregory M. Shaver for their generous inputs as my advisory committee members. Special thanks are due to Dr. J. Stuart Bolton for his valuable advice and support on acoustic measurements. My sincere thanks are due to Robert Brown in the technical shop at Herrick Laboratories, for his great support in fixing and maintaining the engine setup for my project. I wish to express my great appreciation to Bert Gramelspacher in the electronic shop for design and rebuild of the driver unit for fuel injectors. I would also like to extend my thanks to Ronald Evans, David Meyer and Kwok Lee at Herrick for their help during various stages of putting the engine back to work. I wish to acknowledge, in particular, the precious help and valuable input from Jinpu Yang and Chaitanya Shankar Bhat, previous students working on this project. I also wish to thank Seungkyu Lee, a PhD student at Herrick, for his help and advice in preparation for acoustic measurements. I would like to thank US Army Power Generation Branch for initial funding of this project. I would also like to thank Scheid Diesel Services and Bosch for their help and support on rebuilding the injectors. Finally, I wish to thank my parents for their support in my pursuit of graduate study at Purdue, without which this work could not have been completed in the first place.

5 iii TABLE OF CONTENTS Page LIST OF TABLES... iv LIST OF FIGURES... v ABSTRACT... vii CHAPTER 1. INTRODUCTION Motivation Objectives Organization... 3 CHAPTER 2. BACKGROUND AND LITERATURE REVIEW Electronic Diesel Injection Systems Noise from Diesel Engines Sources of Noise from Diesel Engines Combustion-Induced Noise Structural Attenuation from Engine Block... 8 CHAPTER 3. EXPERIMENTAL SETUP Common-Rail-Based Electronic Injection System Acoustic Measurement CHAPTER 4. ENGINE BLOCK STRUCTURAL ATTENUATION MODEL Assumptions of Engine Block Attenuation Attenuation Results from Experimental Measurements Modeling and Simulation Averaged Magnitude-Frequency Response and Estimated Transfer Function Spring-Mass-Damper Model Model Parameter Estimation Model Validation CHAPTER 5. SOUND POWER PREDICTIONS Sound Power Predictions by Applying Experimental Attenuation Curves Sound Power Predictions by Simulation with Nonlinear Models CHAPTER 6. CONCLUSIONS Summary of Contributions Summary of Results Recommendations for Future Work LIST OF REFERENCES... 54

6 iv LIST OF TABLES Table Page 2.1. Representative Outdoor Noise Level Regulations [4] Coordinates of Microphone Locations for Sound Pressure Measurement (Reproduced from [2]) Root Mean Square Errors of Simulated Attenuation Curves with Different Model Parameters (Idle) Root Mean Square Errors of Simulated Attenuation Curves with Different Model Parameters (Full Load)

7 v LIST OF FIGURES Figure Page 1.1. Representative Noise Attenuation System of an Engine Block Attenuation Curves with Various Single Pulse Injection Timings for a Diesel Engine at Idle (left) and Full Load (right) [2] Cylinder Pressure, rate of pressure rise and acceleration of pressure rise of a high performance diesel engine [9] Cylinder pressure diagrams and spectra, noise spectra of a D.I. diesel engine at 1000 rpm full load [12] Structural and acoustical attenuation of a diesel engine [13] Block attenuation curve: estimated in [15] (-) and proposed in [14] (--) Schematic of a Bosch Common-rail Fuel Injection System [17] Solenoid Current Trace for CRIN3 Injector (Courtesy: Bosch) Schematic of Control System Signal Flow (Reproduced from [1]) Sound pressure measurement setup in hemi-anechoic chamber [2] Locations of Microphones for Sound Pressure Measurement [2] Schematic of Engine Block Attenuation System Attenuation Curves for the Engine Block with Various Injection Timings for Single Injection at Idle Speed [2] Attenuation Curves for Engine Block with Various Injection Timings for Single Injection at Full Load [2] Magnitude-Frequency Responses of the Engine Block Attenuation System at Idle Speed (Blue: Averaged Experimental Attenuation Curves; Green: Simulated Attenuation Curves from Transfer Function in Equation (4.1)) Schematic of a Spring-Mass-Damper System Simulink Block Diagrams of Mass-Spring-Damper Model Experimental Cumulative Cylinder Pressure for Various Advanced Crank Angles of a Single Injection Pulse (Idle Speed) Comparison of Experimental Attenuation (top) and Simulated Attenuation (bottom) at Idle Speed (Linear Model) Comparison of Experimental Attenuation (top) and Simulated Attenuation (bottom) at Idle Speed (Nonlinear Damper-Only Model) Comparison of Experimental Attenuation (top) and Simulated Attenuation (bottom) at Idle Speed (Optimized Nonlinear Spring and Damper Model) Absolute Averaged Errors between Simulated and Experimental Attenuation Curves (Idle Speed) Experimental Cumulative Cylinder Pressure for Various Advanced Crank Angles of a Single Injection Pulse (Full Load) Comparison of Experimental Attenuation (top) and Simulated Attenuation (bottom) with Full Load (Linear Model)

8 vi Figure Page Comparison of Experimental Attenuation (top) and Simulated Attenuation (bottom) with Full Load (Nonlinear Damper-Only Model) Comparison of Experimental Attenuation (top) and Simulated Attenuation (bottom) with Full Load (Optimized Nonlinear Spring and Damper Model) Absolute Averaged Errors between Simulated and Experimental Attenuation Curves (Full Load) Cumulative Sound Power Level (A) Prediction through Experimental Attenuation Curves at Idle Speed Cumulative Sound Power Level (A) Prediction through Experimental Attenuation Curves at Full Load Cumulative Cylinder Pressure Used for Sound Power Predictions in This Work (left) and for Calculations of Experimental Attenuation Curves from [2] (right) (Idle Speed) Cumulative Cylinder Pressure Used for Sound Power Predictions in This Work (left) and for Calculations of Experimental Attenuation Curves from [2] (right) (Full Load) Cumulative Sound Power Level Prediction by a Linear Model (Idle Speed) Cumulative Sound Power Level Prediction by a Nonlinear Damper-Only Model (Idle Speed) Cumulative Sound Power Level Prediction by a Nonlinear Spring and Damper Model (Idle Speed) Cumulative Sound Power Level Prediction by a Linear Model (Full Load) Cumulative Sound Power Level Prediction by a Nonlinear Damper-Only Model (Full Load) Cumulative Sound Power Level Prediction by a Nonlinear Spring and Damper Model (Full Load)

9 vii ABSTRACT Cao, Jiajun M.S.M.E., Purdue University, August Analysis and Simulation of Nonlinearities in Noise Attenuation Model for a Diesel Engine Block. Major Professor: Dr. Peter H. Meckl, School of Mechanical Engineering. The purpose of this thesis is to understand the characteristics of noise attenuation of a diesel engine block, by means of setting up a structural attenuation model and exploring the impact of variations of the model, in terms of nonlinearities. A model motivated by a set of experimentally-determined attenuation measurements was constructed, validated and explored for the potential for sound power prediction, with the aid of simulation. Attenuation curves have long been considered an effective and straightforward method to understand the relationship between cylinder pressure and the corresponding noise radiation from an engine block. Preliminary measurements on a small single-cylinder diesel engine, however, suggest dependency of attenuation curves on injection parameters and operating conditions of the engine. Such dependency signals a possibility of inherent nonlinearities in the engine block attenuation and calls for a deeper investigation of the hypothesis. A mass-spring-damper-based model was developed from the averaged experimental attenuation measurements and served as an alternative to attenuation curves to represent the noise attenuation characteristics of the engine block. The hypothesis of inherent nonlinearities was tested with the model and simulation results demonstrated relevance of nonlinearities to several off-linear behaviors of experimentally-determined attenuation curves. Validation for the model was performed by simulation under different operating conditions, and consistent observations further justified the nonlinear hypothesis. Nonlinearities were also categorized to account for different behaviors of the attenuation model. Based on the model developed, sound power levels were predicted for a given set of cylinder pressures under different operating conditions. The simulation-model-based predictions demonstrated more reliability compared to the attenuation-curves-based prediction. Predictions among different models, linear or nonlinear, were compared and the impact of nonlinearities was analyzed. Judging the validity and accuracy of the predictions, however, requires further experimental measurements in the future.

10 1 CHAPTER 1. INTRODUCTION Diesel engines are widely used in industry due to their high thermal efficiency and potential for increased power output when compared with other internal combustion engines such as gasoline engines. Unlike gasoline engines, where combustion starts with the ignition of air-fuel mixture by a spark, diesel engines feature compression-ignited combustion, where air is compressed in cylinder so greatly that the temperature exceeds the ignition point of fuel and subsequent fuel injection starts the combustion process. Such design requires a very high compression ratio, which leads to the most notable advantage of diesel engines increased thermal efficiency. While a high compression ratio does increase efficiency and fuel economy, the great variations of pressure in the combustion chamber also bring a significant amount of noise. Many studies and analyses have been performed on identifying the sources of noise and in particular the correlation between the combustion process and combustion-induced noise. This thesis mainly focuses on the study of the relationship between cylinder pressure, which gives a quantitative description of the combustion process, and the radiated noise pressure level from the engine. 1.1 Motivation The relationship between cylinder pressure and radiated noise level is considered as a structural attenuation from the engine block, as illustrated in Figure 1.1. It is a characteristic of the engine block and is independent of operating conditions such as engine speed, applied load and injection strategies. With a reasonable model, it is possible to predict noise level under certain operating conditions based on the cylinder pressure measured. Operating Conditions Input Cylinder Pressure Engine Block Attenuation System Output Sound Pressure Noise Power Figure 1.1. Representative Noise Attenuation System of an Engine Block.

11 2 Previous studies mainly adopted a linear approach to understand the attenuation and represented the system with an attenuation curve computed from the differences of spectra of cylinder pressure and radiated noise pressure level. Ideally, the attenuation curve is unique for a particular engine and independent of operating conditions, if the linear assumption holds true. Consequently, with the help of the one and only attenuation curve, noise pressure can be readily predicted under any circumstances provided that the corresponding cylinder pressure measurements are available. In an attempt to predict and control noise radiation, a 3kW Tactical Quiet Generator (TQG) MEP 831A with a Yanmar L70-AE DEG FRYC 296cc single-cylinder, naturally-aspirated direct-injection engine was used for experimental study. During the earlier stage of the project, the fuel injection system, originally mechanical with fixed injection timing at 17.0 ± 1 BTDC, was replaced with electronic injection with dspace DS1103 as the Engine Control Unit (ECU) to provide the capability of controlling injection parameters, thus varying cylinder pressure [1]. Measurements of cylinder pressure and corresponding sound pressure of noise radiation were then conducted [2], with various injection timings and different speeds and loads, to compute the attenuation curve for the engine block. The preliminary results, however, demonstrated the dependency of attenuation curve on injection parameters (Figure 1.2). The differences between attenuation curves generated under different conditions can be as large as 10 db, and hence multiple attenuation curves, instead of one, are needed to accurately describe the engine block attenuation system. A possible explanation for these results is that the engine block attenuation system is essentially nonlinear, which also means the attenuation curve might not be the best choice to model the system. As a result, a nonlinear model is proposed in this thesis to analyze the hypothesis of inherent nonlinearities and to better understand the engine block attenuation. Figure 1.2. Attenuation Curves with Various Single Pulse Injection Timings for a Diesel Engine at Idle (left) and Full Load (right): mi stands for main injection, and the number that follows stands for advanced crank angle to top dead center when injection takes place, e.g. mi12 stands for a single pulse injection, which is naturally the main injection pulse, at 12 crank angle before top dead center [2].

12 3 1.2 Objectives This study mainly focuses on a nonlinear structural attenuation model modified from a traditional linear approach for a single-cylinder diesel engine. This thesis started from the experimental cylinder pressure and noise radiation data collected by the previous researcher [2], as well as the corresponding attenuation curves, generated a data-driven model for simulation and added in different kinds of nonlinearities to understand the attenuation system. The study aims to identify the nature of possible nonlinearities in the engine block attenuation system, as well as to explore the potential for prediction of noise level based on the model proposed with only cylinder pressure measurements. 1.3 Organization This thesis is divided into six chapters. Chapter 1 provides the motivation and general background of this study as well as main objectives and a quick overview of the whole thesis. Chapter 2 focuses on more details of the background of the topic and a literature review of relative and important concepts and results shared by other previous researchers in the area. This includes an overview of diesel engines and electronic injection systems, and analysis of noise from diesel engines. Chapter 3 provides a description of the hardware setup on which the measurements used in this thesis were recorded, including an introduction to the generator and the engine, its modified electronic injection system and control unit, and the procedures for acoustic measurements. Chapter 4 focuses on the engine block structural attenuation model, proposes a model and explores the impact of adding nonlinearities. Different kinds of nonlinearities are discussed and the simulation results are compared with the experimental measurements. Several models are generated and validated through predictions of attenuation under different operating conditions. Chapter 5 explores the potential for using the simulated attenuation model to predict sound power radiation from the engine and the influence of nonlinearities on predictions. Finally, Chapter 6 serves as a summary of the work included in this thesis and provides recommendations for future work.

13 4 CHAPTER 2. BACKGROUND AND LITERATURE REVIEW 2.1 Electronic Diesel Injection Systems Conventional diesel engines use mechanical injection systems where the fuel injection is driven by the engine crankshaft. During the compression stroke, the high-pressure pump, which delivers fuel into the injectors, is triggered at some predetermined crank angles in advance of top dead center (TDC). The high-pressure fuel then pushes open the injector and is injected into the combustion chamber. Such design ensures precise and consistent timing of start and duration of injection for each combustion cycle. However, research results show that the injection timing plays a vital role in fuel economy and emissions and consequently calls for a variable and controllable injection. An electronic injection system, where an electromagnetic solenoid actuator controls injection and encoders on the engine crankshaft return real-time crank angle readings, provides the capability for researchers and engineers to vary injection strategies and optimize the performance including efficiency and emissions. Although electronic injection systems have already been widely used in gasoline engines in the 1980 s, the progress of development of electronic diesel injection was relatively slow, mainly due to failure of early electromechanical systems to meet the requirements of high pressure and precise injection in D.I. diesel engines. The development of the diesel injection system started from only adopting electronic sensors and actuators while keeping all basic hydro-mechanical elements, to digital jerk-pump-based injection and finally progressed to modern common-railbased electronic injection system [3]. Apart from that, modern engines are also equipped with digital sensors to measure air flows, combustion chamber pressure, temperature, exhaust emissions and other quantities that indicate the performance of engines. Such measurements, when combined with electronic injection systems, can be built into engine management and control systems where performance issues such as cold start, fuel consumption, chemical and acoustic emissions can be optimized.

14 5 2.2 Noise from Diesel Engines Diesel engines bring bigger problems when it comes to acoustical emission due to greater pressure variations when compared with gasoline engines. Whether the engines are for on-road purposes such as in trucks and trailers, or for non-road applications such as in backhoes or generators, the noise is generally undesirable not only because it causes discomfort for the operators but also because it violates laws and regulations which get stricter every year. Table 2.1 shows representative regulations on outdoor noise levels. Table 2.1. Representative Outdoor Noise Level Regulations [4] Sources of Noise from Diesel Engines Before making any attempt to reduce the noise radiated from engines, it is important to identify the sources of noise first. Noise from diesel engines is caused by vibration of surfaces of structures, attached accessories and covers. Such vibration is mainly attributed to two basic forces: combustion forces resulting from variation of cylinder pressure in the combustion chamber and mechanical forces from engine mechanisms such as crankshaft, gear trains and pistons [5]. Austen [6] in his work identified 6 sources of noise from diesel engines: 1) Fuel Injection Equipment: The noise from injection equipment is low in intensity compared with engine noise but more observable at low engine speed and on small engines. 2) Pump Noise: The noise intensity depends greatly on mounting. A rigid mounting of pumps and flywheel, if there is any, keeps such noise negligible compared to noise from other sources. 3) Injector Noise: The noise mainly comes from the nozzle spring surge and the impact of nozzle needle with the seat on closing. Both sources can be significantly reduced by minimizing the needle mass and spring force. With a proper design, the injector noise could be negligible. 4) Engine Covers: This is a major source of noise including noises from valve cover and timing cover. It is possible to reduce such noise by replacing light cast aluminum covers with heavier steel covers reinforced with rubber or asbestos wool.

15 6 5) Flywheel and Front Pulley: The noise could be radiated at high frequencies but the noise from the flywheel, which is significant, will only be a problem if the engine is tested without a clutch housing. 6) Basic Engine Noise: This is the noise resulting from cylinder pressure variations, which in turn excites other sources of noise mentioned previously. The noise is radiated from the main surfaces of the engine structure, often the most significant and directly related to the combustion process. Since the combustion process also plays an important role in other performance attributes of engines, the basic engine noise is of most interest to researchers and also the focus of this thesis Combustion-Induced Noise Noise from vibrations caused by mechanical forces is generally negligible or independent of operating conditions of engines. Acoustical insulation is often applied to reduce such noise. Combustion-induced noise, or the basic engine noise discussed in the previous section, is a major contributor of the overall acoustical emission and is directly dependent on operating conditions such as speed, load and injection strategies. Such noise sources not only include those directly from the vibrations of the engine main structure due to variations of cylinder pressure, but also those from vibrations of any other components and accessories because of excitation of the combustion forces and sometimes even coupled with the mechanical forces. Engine noise can certainly be reduced by conventional acoustical methods including acoustic barriers, acoustic insulation, isolation mounts, and exhaust silencers, maximizing distance and even placing the whole engine in a sound-attenuating enclosure [6]. However, study of the correlation between noise and combustion helps understand the nature and components of noise radiated, which gives guidelines in noise reduction strategies, and also shows the potential to reduce noise from the source. Ricardo [7-8] in the 1930 s first recognized the relationship between engine noise and rapidity of combustion and detonation, and described it as explosion strikes the walls of the cylinder with a hammer blow. The rapidity of combustion and detonation is perfectly described by cylinder pressure and many studies followed to find the correlation between cylinder pressure and noise from the engine. The shape of the cylinder pressure curve was analyzed, and criteria were generated to describe it [9]: 1) Peak cylinder pressure 2) Rate of pressure rise 3) Acceleration of pressure rise

16 7 Figure 2.1. Cylinder Pressure, rate of pressure rise and acceleration of pressure rise of a high performance diesel engine [9]. Many early investigators tried to understand the combustion-induced noise, both theoretically and experimentally, based on these three criteria. Some argue the rate of pressure rise is a good indicator of noise level, some argue the acceleration to be a decisive factor while the others argue the peak cylinder pressure and the duration it lasts predominates the noise intensity. In the 1940 s, Withrow, Fry and Stone [10-11] drew a very important conclusion that the correlation between the shape of cylinder pressure only exists at low engine speeds while at high speeds it is the peak cylinder pressure that is dominant. It shows that investigation with only one of the three criteria is incomplete, and due to the complexity of looking at all three criteria at the same time, another approach is needed to analyze the cylinder pressure. Thanks to the development of spectrum analysis, researchers started to look at the cylinder pressure in the frequency domain where the information from all three criteria is included. Priede [12] in 1956 plotted the cylinder pressure spectra from a D.I. diesel engine with different injection timings (Figure 2.2), marking the start of investigation of cylinder pressure and noise

17 8 level spectrum. The differences of the cylinder pressure level spectrum and noise level spectrum are considered a characteristic of the sound-attenuation of the engine block and defined as the attenuation curve. Figure 2.2. Cylinder pressure diagrams and spectra, noise spectra of a D.I. diesel engine at 1000 rpm full load [12] Structural Attenuation from Engine Block The structural attenuation is an attenuation spectrum curve and is expected to be independent of operating conditions such as speeds, loads and injection strategies. The concept, which fully describes the acoustical characteristic of an engine block, was proposed by Austen and Priede in 1958 [13]: Attenuation (db) = cylinder pressure (db) sound pressure level (db) (2.1) Figure 2.3 shows the attenuation curve estimated by Austen and Priede for a diesel engine, which demonstrates independency of operating conditions. Such linear attenuation model is

18 9 easy to compute and gives a good measure of sound-attenuating capability of the engine block or any enclosures it is placed within. Anderton [14] and Desantes [15] also generated attenuation curves based on this linear model (Figure 2.4). Figure 2.3. Structural and acoustical attenuation of a diesel engine [13]. Figure 2.4. Block attenuation curve: estimated in [15] (-) and proposed in [14] (--). This technique, however, doesn t work in all cases and Priede later showed that on engines with smooth cylinder pressure profiles, considerable discrepancies and inaccuracies were observed [16]. He first explained that the mechanical-induced noise needs to be separated in order to

19 achieve a linear relationship between pure combustion-induced noise and cylinder pressure level. However he later realized the mechanical-induced noise is coupled with cylinder pressure and could be possibly increased as cylinder pressure level increases. In order to isolate combustion-induced noise, more sophisticated design was added to the test [9]. 10

20 11 CHAPTER 3. EXPERIMENTAL SETUP The subject studied for noise attenuation is a Tactical Quiet Generator (TQG) MEP 831A with 3kW rated power. The generator is powered by a Yanmar L70-AE DEG FRYC 296cc single cylinder diesel engine. As a small engine, it was originally equipped with mechanical injection system and speed governors. In order to provide flexibility of injection strategies and study the relation between injection parameters and noise level, an electronic common-rail-based injection system was developed in the previous phase of the project [1]. 3.1 Common-Rail-Based Electronic Injection System Instead of low-pressure pumps delivering fuel to each injector, a high-pressure fuel rail is used for all injectors with solenoid valves to control fuel delivery. The rail is named as common because it generates the same pressure and serves as the same source of fuel to all injectors on an engine. Figure 3.1 illustrates a typical common-rail-based electronic fuel injection system. The fuel is filtered first and then pumped to high pressure and subsequently delivered to the common rail. A pressure-limiting valve and rail pressure sensor are mounted on the fuel rail. The pressure-limiting valve is usually designed as an electronic regulator, which opens up when the rail pressure exceeds the limit pre-determined by electric signals, resulting in the excessive fuel returning back to the fuel tank. The rail pressure is then roughly maintained at some certain level. However, for a more precise control of rail pressure, a rail pressure sensor needs to be incorporated. Along with the electronic control unit (ECU), which processes signal and control algorithms, a closed-loop control of rail pressure can be realized. One of the major advantages of common-rail injection is that it allows for multiple injection strategies. As opposed to single injection, multiple injection may include pilot injection, which reduces noise level, main injection and post injection, which reduces exhaust emissions. Consequently, injection strategies are more flexible since not only main injection timing but also pilot/post injection quantity and delays between multiple injections can be varied. Since the subject in this thesis is a single-cylinder engine, only one injector is connected to the common rail. The other outlets of the rail are plugged and rail pressure sensor and pressure regulator are still incorporated to control and maintain the rail pressure. The injector used in this project is Bosch CRIN3 and the solenoid current trace is illustrated in Figure 3.2. The energizing time (ET) is a variable to control the duration of the opening of the solenoid valve of the injector. With the current rail pressure reading and fuel quantity, which is readily available on the fuel map, the appropriate energizing time can be estimated.

21 12 Figure 3.1. Schematic of a Bosch Common-rail Fuel Injection System [17]. Figure 3.2. Solenoid Current Trace for CRIN3 Injector (Courtesy: Bosch).

22 13 Figure 3.3. Schematic of Control System Signal Flow (Reproduced from [1]). The electronic control unit used in this project is dspace DS1103 control board. It not only works with rail pressure sensor and regulator valve, but also communicates with the encoder that measures crank angle, and the pressure transducer, which reads the cylinder pressure. An illustrative scheme is shown in Figure 3.3, demonstrating how signals flow in the whole system. All measurements and control signals are crank-angle-based but the time vector is also recorded by dspace [1]. 3.2 Acoustic Measurement Noise from engines needs to be accurately measured in order to indicate the power level of acoustical emission. Ideally, engines should be placed in a free-field environment where there are no obstacles or reflecting surfaces, which generate stationary acoustic waves, and measurements should be taken over the whole surface of a virtual enclosure. However such a perfect free-field environment is not always practical and it is acceptable in engineering to take measurements in an essentially free-field over a reflecting surface, according to ISO 3744.

23 14 Figure 3.4. Sound pressure measurement setup in hemi-anechoic chamber [2]. Acoustic measurements generally start with a microphone that precisely measures the pressure variations around ambient atmospheric pressure and returns the quantity as a voltage. Power supplies and signal conditioners are often needed to supply current, polarize transducers, amplify charges and filter signal noise. In practice, a hemi-enclosure, rectangular or spherical, is used as surface for measurement. Figure 3.4 shows the acoustic measurement setup for this project. The engine was placed in a hemi-anechoic chamber where walls are coated with high sound-absorption foams but an acoustic reflecting floor plane is allowed. Such setup complies with the essential free-field environment described in ISO 3744 and is appropriate for acoustic measurement.

24 15 Figure 3.5. Locations of Microphones for Sound Pressure Measurement [2]. The hemisphere enclosure used is 1.75 m in radius and there are 10 microphones placed at different locations. The illustration and coordinates of microphone locations are shown in Figure 3.5 and Table 3.1.

25 16 Table 3.1. Coordinates of Microphone Locations for Sound Pressure Measurement (Reproduced from [2]). Microphone Locations x (m) y (m) z (m) Since the range of sound pressure measurements varies greatly, a logarithm scale is commonly used and defines the sound pressure level: SPL = 10 log( p rms 2 ) db (3.1) p ref 2 where p rms is the effective mean pressure and p ref is the reference pressure, usually taken as the threshold of human hearing capability, 20 μpa. An important rating in acoustics is the sound power. As opposed to sound pressure, which depends on the distance to the source and the properties of the media in which the sound travels, sound power is independent of the surroundings and is an indicator of the total energy emitting from a sound source. The sound power is defined as: W = IdS (3.2) where S is the surface area and I is the acoustic intensity defined as: I = p rms 2 ρ 0 c (3.3) where ρ 0 is the density of air, usually taken as 1.21 kg/cm 3, and c is the speed of sound in air, usually taken as 343 m/s. Sound power is also commonly represented in logarithm scale, and the sound power level is defined as: SWL = 10 log( W W ref ) db (3.4)

26 17 where W ref is the reference sound power, expressed as: W ref = p ref 2 where S ref is taken as 1 m 2, W ref is around W. ρ 0 c S ref (3.5) If the average sound pressure from 10 microphones is used to calculate the sound power in Equation (3.2), the relationship between sound power level and sound pressure level can be found as: SWL = SPL + 10 log( S S ref ) db (3.6) The db scale treats all frequencies equally, which does not reflect the nature of human hearing. Human ears are particularly sensitive to frequencies in a range from 1000 to 4000 Hz. The A- weighting filter approximates the hearing characteristics and was adopted by OSHA in 1972 as the official sound pressure level correction method [6]. In ANSI standards [18], the expression for A-weighting filter is: where 1.562f 4 W A (f) = 10 log [ ] + W (f )(f ) C(f) db (3.7) W C (f) = 10 log [ f 4 (f ) 2 (f ) 2] db (3.8) The A-weighted sound pressure level and sound power level are represented as: SPL A (f) = SPL(f) + W A (f) db(a) (3.9) SWL A (f) = SWL(f) + W A (f) db(a) (3.10) where db(a) indicates a frequency-a-weighting scale.

27 18 CHAPTER 4. ENGINE BLOCK STRUCTURAL ATTENUATION MODEL In an attempt to understand and predict the noise radiated from engines, the structural attenuation, which captures all acoustic characteristics of the engine block, was studied in this thesis. This concept was first proposed by Austen and Priede [13] and then adopted by many researchers and even engine manufacturers as an indicator of engine performance in terms of acoustic emission. The attenuation model, however, was completely based on a linear assumption and any discrepancies during modeling were attributed to the non-negligible mechanically-induced noise, which was not supposed to be included in the linear part of the cylinder pressure noise level relationship. The focus of this thesis is to challenge the linear assumption of the engine block attenuation model and to investigate the impact of different nonlinearities on the behavior of the model. Measurements of cylinder pressure and corresponding noise level under several operating conditions were performed at the previous stage of the project and an attenuation model, represented as attenuation curve over a range of frequencies, was generated via a conventional linear approach. However, attenuation curves generated under different operating conditions were not consistent and therefore calls for study of the attenuation model from another perspective, in a nonlinear way. Just as most nonlinear problems start with a linear approximation, a linear attenuation model that best describes the averaged characteristics of attenuation curves generated in different cases was set up. The model was represented in terms of an attenuation curve, transfer function, differential equations and finally a mass-spring-damper system. The physical system resembles the main characteristics of the original attenuation model, and provides freedom to easily build in nonlinearities such as nonlinear springs or dampers. Finally, through simulation of the mass-spring-damper model, the impact of nonlinearities was studied and compared with a purely linear model. 4.1 Assumptions of Engine Block Attenuation The structural attenuation model of engine blocks comes with several important underlying assumptions. These assumptions lead to a straightforward definition of attenuation (Equation (2.1)) but also introduce inaccuracies in modeling the acoustic features of the engine block and subsequently account for some discrepancies when experimentally measuring and computing the attenuation. The major assumptions include:

28 1) Dominance of combustion-induced noise: The noise radiated from the engine is considered to be mainly, if not wholly, a result of forces excited by combustion. This includes both combustion forces that come directly from the variation of cylinder pressure and mechanical forces that come from the movement of various parts in the engine including piston, gear trains, crankshaft, etc. but only as a result of combustion. Mechanical forces that are independent of the combustion process are generally considered to have little contribution to the overall noise emission and thus are often neglected. Techniques to separate noise induced purely by mechanical forces independent of combustion, such as periodic piston movement or injector valves opening, are sometimes adopted but in most cases overlooked if there is no indication this could be a problem. 2) Cylinder pressure representative of combustion process: This assumption states that cylinder pressure alone is enough for fully describing the combustion process, at least from an acoustic perspective, and can be used solely to compare with the corresponding noise emission and compute the attenuation. While combustion does result in substantial variations in cylinder pressure, change of cylinder pressure cannot always be attributed to combustion. Cyclic piston movement, intake of air and emission of exhaust all contribute to cylinder pressure variations and don't necessarily have connections with a particular combustion process. Variations in cylinder pressure from these sources, however, excite noise from the engine block and consequently the difference of total cylinder pressure and noise radiation level still serves as a valid acoustical structural attenuation model for the engine block. Cylinder pressure decomposition can be realized by subtraction of cylinder pressure when the combustion is not present, usually the free rotating cycles when there is no fuel injection, from the measurements of cycles when fuel injection and subsequent combustion are present. Similar decomposition could also be applied to acoustic measurements of noise and the attenuation model generated would be between only combustion-related pressure variations and combustion-induced noise. 3) Linear assumption of engine block attenuation: The engine block attenuation can be regarded as a system that captures the acoustic characteristics of the engine block and takes cylinder pressure as input and noise radiation level as output. Figure 4.1 shows a schematic of the engine block attenuation system, which essentially attenuates the cylinder pressure inputs and results in sound pressure outputs at a certain distance from the center of the engine block, indicating the level of noise radiation. A linear attenuation model assumes the relationship between cylinder pressure and engine noise radiation to be purely linear. Just as any linear system, such model follows the rule of superposition and preserves frequency of harmonic input. In the frequency domain, the linear assumption indicates that only one attenuation curve independent of input, which reflects different operating conditions of the engine, is allowed since the net attenuation of different harmonic components should only be a function of frequencies. Computation of the difference of spectrum of cylinder pressure and noise radiation level is a purely linear approach and any discrepancies from different computations could possibly result from an invalid linear assumption, in addition to measurement uncertainties and computation errors. 19

29 20 This thesis mainly investigates the validity of the linear assumption but still carries out the remaining analysis based on the first two assumptions. Although the linear assumption is challenged, the linear approach of generating attenuation was still adopted throughout the study with the exception that, this time, discrepancies of attenuation curves from different computations were explained by the introduction of nonlinearities. Cylinder Pressure Input Engine Block Attenuation System Averaged Sound Pressure at 1.75 m away from Center Output Figure 4.1. Schematic of Engine Block Attenuation System. 4.2 Attenuation Results from Experimental Measurements To understand the impact of the combustion process on noise radiation, different injection strategies were tested during the previous phase of the project, among which the most studied was the injection timing of single injection pulses [2]. The injection into a diesel engine happens slightly before the piston hits top dead center (TDC) during the compression stroke of each cycle and the advancing time is measured in crank angles. The injection timing determines the start of combustion and subsequently the cylinder pressure shape and completeness of combustion, which directly relate to efficiency, emission and noise radiation of an engine. Generally speaking, an early injection would result in an early start of combustion, rapid rise in cylinder pressure and temperature, and consequently lead to higher efficiency, decreased emission of soot, but increased NO x exhaust and noise levels. Test runs of five different injection timings each under two different operating conditions in terms of engine speed and load were conducted. The injection timings vary between 12 before top dead center (BTDC) and 24 BTDC, while the operating conditions are idle speed at 3050 rpm and zero load, and full load with 3450 rpm. For each of the 10 cases, both cylinder pressure and sound pressure were measured at the same time in the hemi-anechoic chamber (Figure 3.4). The cylinder pressure was measured by a pressure transducer preinstalled and the sound pressure was measured by 10 microphones at 1.75 m from the center of the engine (Figure 3.5). Attenuation was computed according to Equation (2.1) for each case and the detailed steps include: 1) Generate the 1/3 octave band spectrum of cylinder pressure level over frequencies from 20 to 20,000 Hz from time history of cylinder pressure measurements. 2) Generate the 1/3 octave band spectrum of sound pressure level over frequencies from 20 to 20,000 Hz from the averaged time history of 10 microphone measurements and apply A- weighting filter to the spectrum.

30 21 3) Compute the difference between 1/3 octave band spectra of cylinder pressure and of A- weighted sound pressure level and express the result as an attenuation curve over frequencies from 20 to 20,000 Hz. The resulting attenuation curves of different injection timings under different operating conditions are plotted in Figures 4.2 and 4.3. Figure 4.2. Attenuation Curves for the Engine Block with Various Injection Timings for Single Injection at Idle Speed [2].

31 22 Figure 4.3. Attenuation Curves for Engine Block with Various Injection Timings for Single Injection at Full Load [2]. Significant discrepancies can be observed from these attenuation curves with different injection timings for each operating condition although general trends are similar. Attenuation is relatively low over high frequencies and is least around 5,000 Hz. Lower attenuation means greater contribution of cylinder pressure to the overall noise level and thus it is important to understand the discrepancies in the range of high frequencies where the attenuation is low, by introduction of a new model. Note that although it is more important to focus on the range of high frequencies, the discrepancies among different attenuation curves don't necessarily grow with frequencies. Since the attenuation is represented in a logarithm scale, a fixed difference in db indicates a fixed percentage variation and the absolute differences grow as the base level increases. Hence the discrepancies in low frequency range could be as significant as in high frequency range because the averaged attenuation is much higher in low frequency range than in high frequency range, although in a logarithm scale discrepancies in high frequency range seem dominant. There is a noticeable amplitude shift of attenuation curves for different injection timings in both idle and full load cases, and the shift appears to be in sequence of the advanced timing of injection. Namely, the attenuation of the earliest injection is the highest whilst a late injection accompanies a low attenuation. Since early injection also increases both the peak of cylinder pressure itself and the rise rate, the change of attenuation can be accounted for by change of amplitude of the input to the system, which is cylinder pressure. Qualitatively, larger amplitude of cylinder pressure input results in a higher attenuation. This behavior is very similar to that of a typical nonlinear system and hence a nonlinear model was built to verify the hypothesis.

32 23 There are also other discrepancies among the attenuation curves, although less evident or seemingly of little causality with the injection timings or operating conditions. For instance, a sub-resonant peak around 500 Hz is found only in some cases of different injection timings. To judge whether these behaviors are also the results of nonlinearities in the attenuation model requires the construction of a nonlinear model first. 4.3 Modeling and Simulation Representing the acoustic characteristics of an engine block with an attenuation curve, generated from differences of input cylinder pressure spectrum and output sound pressure level spectrum, is a purely linear approach. In order to incorporate nonlinearities, the attenuation curves need to be transformed into another representation that applies to both linear and nonlinear systems. The most fundamental mathematical representation of a system would be differential equations. The modeling procedure in this thesis started with an averaged linear attenuation curve, estimated the transfer function and converted it into differential equations and an equivalent mass-spring-damper model. A virtual mass-spring-damper model was then developed and simulated with the help of MATLAB SIMULINK Averaged Magnitude-Frequency Response and Estimated Transfer Function The modeling started with a linear approximation that essentially takes the average of the attenuation curves corresponding to various injection timings. The operating condition used for modeling is idle speed and the full load condition is reserved for model validation purpose. The attenuation curve is essentially a mirror image of the conventional magnitude-frequency response curve of a system with respect to the 0 db axis, and a transfer function can be readily estimated to match the general trends of the curve. Approximately speaking, the averaged attenuation decreases at a rate of 40 db/decade until reaching its minimum around 5,000 Hz and increases afterwards at the same rate. This resembles a 4 th order system with 2 zeros. The system can be decomposed into 2 second-order systems each representing the magnitudefrequency response of cylinder pressure and averaged sound pressure, and the A-weighting filter, respectively. Since A-weighting filter is well-defined and can be readily imposed during spectrum analysis, it is temporarily taken out to lower the degree and complexity of the engine block attenuation system. The simplified engine block attenuation system (Figure 4.1) takes cylinder pressure as input and averaged noise pressure measured at a set distance (1.75 m) from the center of the engine as output. The resulting averaged magnitude response when the engine is at idle is plotted in Figure 4.4 as the blue curve. Compared to the attenuation curves in Figure 4.2, the magnitude response curve changes at a rate of ±20 db/decade instead of ±40 db/decade, and there is a break frequency around 100 Hz in addition to the resonant frequency around 5,000 Hz. The differences can be accounted for by taking out the A-weighting filter, which is essentially a 2 nd order system.